Fig 1.
736 patient records were screened for inclusion, 56 records met all study criteria. Sequential definition of the cohort described on the left, methods for defining the cohort by study phase summarized on the right.
Fig 2.
Impact of different definitions for the baseline time period on variable availability.
These results correspond to the 183 patients who were not excluded by non-CDAI related study criteria. All of these patient records underwent manual review to ascertain the presence of PROs occurring during the 16 week period prior to the first dose of ustekinumab (week 0). The blue dotted line corresponds to the cut point (12 week lookback) that was selected following a visual review of the trends depicted here. CRP corresponds to c-reactive protein.
Fig 3.
Comparison of total CDAI as computed using CDAI elements ascertained informatically vs manually.
The y-axis corresponds to the CDAI calculated by database query of non-PRO3 elements. The x-axis corresponds to the CDAI calculated by manual abstraction.
Table 1.
Comparison of total CDAI as computed using CDAI elements ascertained informatically vs manually.
All results correspond to the results of manual abstraction by one chart reviewer (45 patients).
Table 2.
Comparison of the results of non-PRO3, binary CDAI elements by manual vs informatics methods.
Accuracy is reported relative to manually abstracted data (gold-standard). W12 and W24 correspond to the week 12 and 24 periods respectively. Comparisons were made against the annotations performed by one chart reviewer (45 patients), roughly a quarter of the 183 patients who met all eligibility criteria prior to application of the baseline CDAI requirement.
Table 3.
Comparison of the results of non-PRO3, continuous CDAI elements by manual vs informatics methods.
Mean absolute error was calculated only for values that were non-missing by both informatics and manual methods. Comparisons were made against the annotations performed by one chart reviewer (45 patients), roughly a quarter of the 183 patients who met all eligibility criteria prior to application of the baseline CDAI requirement.
Table 4.
Characterization of missing PRO3 elements at baseline.
Proportions correspond to the 183 patients who met all eligibility criteria prior to application of the baseline CDAI requirement.
Table 5.
Comparison of the results of two imputation models.
Imputation models were applied to all of the otherwise eligible patients (less the baseline CDAI requirement) that had been assigned to one chart abstractor (45 patients).
Table 6.
Characterization of the study cohort.
*: Reported as median, other continuous variables reported as mean ± standard deviation. Parenthetical modifiers to the disease location elements correspond to the Montreal Classification of Crohn’s disease.
Table 7.
Characterization of missing PRO3 elements across all time periods.
Proportions correspond to all 56 members of the external cohort.
Fig 4.
Change in the Crohn’s Disease Activity Index (CDAI) over time.
Lines correspond to individual patient trajectories, where turquoise corresponds to patients who were biologic-intolerant or refractory (BioIR) and red corresponds to patients who were biologic-naïve. Open circles represent time points where the patient was not receiving oral steroids whereas filled circles represent the use oral steroids at a given time.
Table 8.
Efficacy endpoints.
Table 9.
Steroid use across timepoints.